This example compares what you can learn about chess openings from a game database (with tree
features such as ChessBase™ and Chess Assistant™) to what you can learn with a positional database (Bookup) that contains the
exact same information.
Backsolving is one of the most sophisticated features of Bookup - and one of the least
understood. This article takes a bit of the mystery out of this feature by showing how it
works compared to statistics.
Suppose you have a huge game database.
You search for your favorite White opening and
find 97 grandmaster games with 12.Ng5.
| |
Games |
White Wins |
Draws |
Black Wins |
| 12.Ng5 |
97 |
51.5% |
31.0% |
17.5% |
| 12.Bd2 |
61 |
34.3% |
32.7% |
32.7% |
| 12.h3 |
3 |
33.3% |
33.3% |
33.3% |
The game tree tells you that about 51% of the games were won by White, 31% were drawn
and 17% were won by Black.
Because you play the White side, the move 12.Ng5 starts to look
very attractive...
But wait - you view some of those games and find that grandmasters have been
playing 12.Bd2 for the past three years instead of 12.Ng5, even though 12.Bd2 shows only a
34% winning percentage for White.
What's going on?
You then look further into the games and find that in just two key games Black has found a
way to completely neutralize 12.Ng5 - and even get a slight advantage!
Naturally after
those games were published, grandmasters gave up on Ng5 for White.
So the game tree
statistics would have lured you into a bad position.
Backsolving does this research for you and makes it obvious when game statistics and
current chess theory contradict each other. Here's what the candidates might look like
after importing those grandmaster games into Bookup 2000 Professional and then backsolving:
The backsolving process was able to show that with best play by both sides, Black could
achieve a slight advantage if White plays Ng5.
Backsolving
This book was also analyzed by a playing program and the score
of -131 means the computer engine agreed with this assessment. Down the 12.Ng5 line it is now
easy to see the Black moves that keep the edge.
Automatic solving is easy; simply change the assessment of the last position in
any line (a 'leaf node') and press the rewind button. As the moves are taken back, each
position is re-evaluated. Bookup 2000 Professional supports two assessment symbols, one for the opinion
of the author and the other for the results of Backsolving.
Backsolving also works with computer evaluations. Combined with overnight
analysis, this allows you to create a theory machine that points out
flaws in your repertoire - and discovers new improved lines.
Backsolving can also show you where the bulk of the positions are in the tree of
analysis. Looking at the candidates, 202 positions are recorded beyond 12.Ng5
while only 17 positions are stored after 12.h3. Of course, even though more analysis
appears after 12.Ng5, it is no longer the main line - White would rather play a move that
leads to an unclear or equal position.
It is precisely
because 12.Ng5 is refuted and is no longer being played that game statistics for that
move are likely to stay the same in the future.
Thus a bad move will continue to look attractive
(statistically) in a game database long after it has been refuted.
Backsolving is a
feature unique to Bookup 2000 Professional, one of
many power tools that make it the premiere software for studying chess openings.
Did we enlighten you? Confuse you? Email us
with your comments. Thanks!
Backsolving
Roman's